84 the perspective of the disaggregated regulatory approach with respect to the logical layer is that there are no network elements which can substantiate market power. There is no justification for regulation on the logical layer of Internet service provision. 5.5 Analyzing the cost characteristics of the physical layer of the Internet Network economics shows that monopolistic bottleneck network areas are most likely to be found on the physical layer of network service provision. The physical layer by far encompasses the largest investments necessary for Internet service provision. This is not because of high costs for purchasing transmission lines, but rather because of the substantial expenditures for the first-time installation of these lines. The large investment cost associated with building a network infrastructure can give rise to economies of scale and scope in the market. Furthermore, the installation costs, such as digging trenches and laying cable, are sunk upon entering the market. The physical layer therefore has a high potential of substantiating stable market power. Figure 5.1 illustrates the physical layer of Internet service provision. Figure 5.1: Stylized illustration of the physical layer Point of Presence: Access point to the ISPs network Point of Interconnection between the networks of the incumbent carrier and the ISP Long-distance communications infrastructure of the ISP PoP POI En d U se rs L a rg e cu st o m er Retail leased lines PoP POI Local loops Incumbent‘s long-distance comm. infrastr. Collocation sites 85 Local communications infrastructure To connect the end-user’s terminal equipment with the ISP’s network, the ISP needs to provide for a physical path between the end-user site and its backbone network. Early in the Internet history is was most common for end-users to use a computer modem for placing an ordinary telephone call to a local access number of an ISP. The call would connect the end-user site with a PoP of the ISP’s network on the basis of local telecommunications infrastructure. From the PoP, the ISP could take over the transmission of the data, convert the signals of the analog telephone call into digitized IP format, and forward the IP packets to its packet-switched IP network. This form of dial-up access has mostly been replaced by broadband access technologies, which offer the end-user far higher download rates from the Internet. For broadband access to the Internet the end-user site needs to be connected with the ISP’s network by a high-capacity link. On the basis of telecommunications transmission lines this is most often accomplished by DSL technology. The frequencies of the phone line are divided into a high-frequency band and a low-frequency band for Internet and voice traffic respectively. The Internet data transmitted over the highfrequency band are handed over to the ISP at a point of interconnection (POI) between the telephone carrier’s network and the ISP’s network. The ISP can aggregate the traffic of several DSL72 customers at the POI and forward the traffic via a leased line or own transmission lines to its backbone network. For larger customers that generate more Internet traffic it can be economical for the ISP to build out new fiber cable to the end-user site or at least to connect the end-user site directly via leased line to its backbone network. The substantial installation costs involved in laying local communications infrastructure can be divided among only a few end-users attached to the same local infrastructure. Local lines are therefore characterized by substantial sunk investments per user. Traditionally, the local loop and local leased lines were considered monopolistic bottlenecks for the provision of voice communication and Internet services. Economies of scale are generally not exhausted in the local network such that the duplication of investments into local communications infrastructure is not economically efficient. Recently, however, the view that all local transmission lines are monopolistic bottlenecks has been put into a new perspective by the trend of convergence between media platforms. The duplication of existing local telecommunications infrastructure is indeed prohibitively expensive, as long as the economies of scale in the local infrastructure are not exhausted and the building of alternative local infrastructures would require incurring substantial sunk investments. However, the investments into access infrastructure are significantly lower when existing fixed-line local infrastructures are upgraded for the use in local communications services. Internationally, 72 DSL stands for Digital Subscriber Line. DSL is a technology that provides digital data transmission over the wires of a local telephone network and thereby realizes higher download speeds. 86 especially the upgrading of Cable-TV lines (cable-modem access) has been highly successful.73 Technological advances in wireless access technologies have further expanded the list of possible substitutes in the local access market which require comparably low sunk investments in network infrastructure. The currently most established wireless access technologies are Wireless LAN (W-LAN) and Worldwide Interoperability for Microwave Access (WiMAX). Büllingen and Stamm (2007: 15ff.) discuss the technical and economical potentials of these technologies. The most important technical drawback of W-LAN and WiMAX is the fact that the offered download rates are shared by the number of users in the same access cell. For this reason, the realized download speed is generally far below DSL standards. The subscription prices for W-LAN and WiMAX are, however, well above the price of DSL or cable-modem services in metropolitan areas. They can therefore not compete with DSL and cable-modem access directly. Only in network areas in which DSL or Cable-access technologies are not available, foremost in rural areas, do wireless technologies have a chance at commercial success. In these instances, they need to offer an adequate service at a lower price than would be charged if a DSL or Cableaccess technology were deployed in this region. A second trend that is dissolving the monopolistic bottleneck in the local access market is the demand for very high bandwidth connections to the Internet. To meet this demand, carriers are increasingly investing in the deployment of new fiber lines closer to end-user sites. Fiber-to-the-home (FTTH) or fiber-to-the-curb (FTTC) projects are especially prevalent in metropolitan areas. These projects also put competitive pressure on existing infrastructures. In the face of new alternative technologies for local Internet access (and also voice telecommunications services), the view that local loops are generally monopolistic bottlenecks can no longer be adhered to. Active or potential competition in the local loop is very selective (Woroch, 2002). The institutional framework (regulatory environment), technological developments, historical circumstances in which alternative technologies have evolved, as well as factors such as the population density, all influence the chances of infrastructure-based competition in a particular local network area. For this reason, it is important that regulation not hamper the incentives to build out alternative infrastructure whenever this is economically efficient. For the purposes of the present analysis, it can be concluded that when economies of scale and scope are not exhausted in the local network area and when offering alternative access technologies involves making substantial sunk investments, then the incumbent telecommunications carrier continues to have market power. Sectorspecific regulation is justified in these instances. The regulation of local communi- 73 In Germany, cable-access is only recently gaining market share in telecommunications and Internet services. This is due to several regulatory and institutional barriers, which delayed investments into upgrading the cable infrastructure. Büllingen et al. (2007) give an overview of current prospects of cable-access in Germany. 87 cations infrastructure needs to be reevaluated regularly, since technological advances are likely to further expand the realm in which competition is possible. Long-distance communications infrastructure As with local communications infrastructure, the installment of long-distance communications infrastructure also involves large fixed investments. Many of these investments are sunk, such as acquiring rights of way, or labor costs and machinery costs for burying cable in the ground. Once a network is in place, the variable costs of operation are relatively small compared to the fixed costs of building a network. This constellation leads to economies of scale. Also, since long-distance communications lines can be used for more than one kind of service (for instance voice telephony and data transmission), they are , in addition, characterized by economies of scope. In the late 1990s the dynamic development of the Internet triggered a surge in demand for long-distance network capacity. At the capacity limit, marginal cost of network use rises sharply as capacity expansion becomes necessary. The economies of scale and scope of long-haul communications infrastructure have, since then, largely been exhausted. Carriers extrapolated the demand growth for capacity, which resulted from the exponential growth rates in Internet usage and invested extensively into long-distance communications infrastructure. Elixmann (2001) gives a comprehensive overview of the international market for long-haul transmission capacity at this time. The study comprises data on the length of transmission lines, investment outlays for network capacity, stock exchange capitalization, turnover, and locations of the most important international carriers for the year 2000. A study by the OECD (2002) shows that many carriers built end-to-end capacity linking the major IXs in the U.S. with their respective home networks. As a result, international IP-traffic today is spread over many networks. Carriers without substantial long-distance communications infrastructure have several options for purchasing transportation capacity on all major routes. The study also describes that carriers engage in “capacity swapping” with international partners. European carriers, for instance, use the network of a U.S.-based partner carrier and vice versa (OECD, 2002: 17). In conclusion, the market for long-distance communications infrastructure is effectively competitive. There are opportunities both for demand-side substitution as well as for supply-side substitution, should one carrier restrict output or raise prices above the competitive level. Regulation of network elements belonging to longdistance communications infrastructure cannot be justified on the basis of the disaggregated regulatory approach. Co-location Co-location space is a physical site at which telecommunications carriers install hardware equipment, such as switches and routers and where network interconnection is realized. Because of the sensitivity of the communications equipment as well as the high security demands for such a site, co-location facilities afford large 88 investments into fire protection, water protection, power supply, air conditioning, and security provisions. The large investment costs for setting-up a co-location site give rise to substantial economies of scale, which only the largest carriers can be expected to exhaust fully. Furthermore, a large amount of the investment costs for building co-location space are sunk. Smaller carriers that require far less space for their equipment and expect to serve fewer customers from a given site are at an economic disadvantage. For these carriers it is efficient to share co-location space and to distribute the overhead costs and sunk costs of co-location among several carriers. Lockable racks can ensure that carrier equipment is protected in shared rooms and cabinets. Whenever ISPs depend on regulated wholesale access to the local access lines of an incumbent telecommunications carrier, they will also need co-location space at which to install their hardware and where to realize the physical interconnection with the incumbent’s network. The amount of space a carrier requires at a given site is a function of the number of users and the amount of traffic volume, which this site serves. For choosing the geographical location of a co-location site it is important to know whether it is likely that several carriers will be present at the prospective site to share alternative co-location facilities. In cosmopolitan areas, where many carriers are in need of co-location space and where a high population density allows these carriers to serve more customers from one collocation site, independent companies that are specialized in offering shared co-location space to carriers, exist.74 Wherever such offers are being made, a carrier can rent co-location space scaled to demand. Compared to the investment costs for building own co-location space, the sunk setup costs involved with using rented co-location space are negligible. The more rural the co-location site is, the lower will be the number of carriers active in the area, and the smaller the number of users that can potentially be served from one site. For this reason, independent organizations do not offer scalable colocation space in rural areas. In these areas, only the incumbent telecommunications carrier will have a high enough demand to justify investments into co-location facilities. Co-location space in rural areas can therefore be a potential monopolistic bottleneck. In these instances sector-specific regulation is justified that requires the incumbent telecommunications carrier to offer wholesale access to its co-location facilities on non-discriminatory terms. Summary physical layer The disaggregated analysis of the physical layer suggests that those network elements which are associated with providing long-distance communications are competitive. The network elements associated with offering access services, such as local loops and co-location space, show different degrees of competition, depending on the population density in a particular region. Local loops are likely to be compe- 74 Companies such as Interxion (www.interxion.com; site last visited on Feb. 15, 2008) and Equinix (www.equinix.com; site last visited on Feb. 15, 2008) for example build and operate Internet exchange centers in large cities worldwide. 89 titive when alternative fixed-line infrastructures are present that can be upgraded to provide Internet services. This is more often the case in metropolitan areas. Alternative co-location facilities are also more likely in metropolitan areas because more carriers are active there, and more potential customers can be served from one particular site. Monopolistic bottlenecks remain in rural network areas. Sector-specific regulation is justified whenever local loops and co-location facilities must be considered monopolistic bottlenecks. In these cases, the incumbent carrier should be obliged to offer non-discriminatory access to its network infrastructure. 5.6 Conclusions Looking at the market for Internet service provision from the viewpoint of the disaggregated regulatory approach, this chapter comes to the conclusion that potential monopolistic bottlenecks can be found only in local communications infrastructure. Local access lines, local leased lines, and co-location space can be associated with economies of scale in the relevant output region and substantial sunk investment costs. These economies of scale have recently been exhausted by upgrading alternative facilities where this was less costly than setting up an entirely new infrastructure. However, especially in rural areas, local communications infrastructures often remain monopolistic bottlenecks. The network elements of the local access infrastructure belong to the physical layer of Internet service provision and are part of the Internet periphery. Sectorspecific regulation already applies to these network elements insofar as they belong to traditional telecommunications markets. Regulation of telecommunications markets has taken into account the importance of non-discriminatory access to local telecommunications infrastructure also for competition in Internet services markets. Chapter 9 provides an overview of telecommunications regulation in the U.S. and in Europe and answers the question whether current regulatory practice is sufficient to guarantee functioning competition in Internet service provision. The disaggregated analysis of the market for Internet services did not disclose any regulatory requirements for long-distance network capacity or for core elements of Internet service provision. These network elements, which are at the center of the discussion on potential Internet regulation, do not show characteristics of monopolistic bottlenecks. From the analysis of the cost characteristics of Internet service provision there is no need for ex-ante sector-specific regulation in long-distance network capacity or in the logical layer of Internet service provision.